Re: [PATCH 2/2] USB: ehci: tegra: Align DMA transfers to 32 bytes

[rmorell@nvidia.com]

On Mon, Jan 24, 2011 at 08:36:55AM -0800, Alan Stern wrote:
> On Thu, 20 Jan 2011, Robert Morell wrote:
> 
> > The Tegra2 USB controller doesn't properly deal with misaligned DMA
> > buffers, causing corruption.  This is especially prevalent with USB
> > network adapters, where skbuff alignment is often in the middle of a
> > 4-byte dword.
> > 
> > To avoid this, allocate a temporary buffer for the DMA if the provided
> > buffer isn't sufficiently aligned.
> > 
> > Signed-off-by: Robert Morell <rmorell@nvidia.com>
> > ---
> >  drivers/usb/host/ehci-tegra.c |   92 +++++++++++++++++++++++++++++++++++++++++
> >  1 files changed, 92 insertions(+), 0 deletions(-)
> > 
> > diff --git a/drivers/usb/host/ehci-tegra.c b/drivers/usb/host/ehci-tegra.c
> > index 2341904..7cdfc65 100644
> > --- a/drivers/usb/host/ehci-tegra.c
> > +++ b/drivers/usb/host/ehci-tegra.c
> > @@ -32,6 +32,8 @@
> >  #define TEGRA_USB_USBMODE_HOST			(3 << 0)
> >  #define TEGRA_USB_PORTSC1_PTC(x)		(((x) & 0xf) << 16)
> >  
> > +#define TEGRA_USB_DMA_ALIGN 32
> > +
> >  struct tegra_ehci_context {
> >  	bool valid;
> >  	u32 command;
> > @@ -461,6 +463,94 @@ static int tegra_ehci_bus_resume(struct usb_hcd *hcd)
> >  }
> >  #endif
> >  
> > +struct temp_buffer {
> > +	void *kmalloc_ptr;
> > +	void *old_xfer_buffer;
> > +	u8 data[0];
> > +};
> > +
> > +static void free_temp_buffer(struct urb *urb, int status)
> > +{
> > +	enum dma_data_direction dir;
> > +	struct temp_buffer *temp;
> > +
> > +	if (!(urb->transfer_flags & URB_ALIGNED_TEMP_BUFFER))
> > +		return;
> > +
> > +	dir = usb_urb_dir_in(urb) ? DMA_FROM_DEVICE : DMA_TO_DEVICE;
> > +
> > +	temp = container_of(urb->transfer_buffer, struct temp_buffer,
> > +			    data);
> > +
> > +	if (dir == DMA_FROM_DEVICE && !status)
> > +		memcpy(temp->old_xfer_buffer, temp->data,
> > +			urb->transfer_buffer_length);
> 
> Even if status is nonzero, there may be valid data in the buffer.  You 
> should skip that test.

Thanks for looking, Alan.  I added that test based on earlier feedback.
I think the big concern here is security: if the URB fails in such a way
that the buffer is not overwritten, then we may copy out freed kernel
data to userspace.

Are there specific status codes that I can check for here?  I guess the
only other option is to remove the direction check from the alloc path
or alloc with GFP_ZERO.

Thanks,
Robert

> No other problems that I can see.
> 
> Alan Stern
> 
>